Remove duplicated release-prefix root mirror files

gwam_v12_sparse_v2/examples/REALTIME_ENV_GRAPH_README.md

@@ -1,90 +0,0 @@
-# GWAM real-time final graph example
-
-Run this only in an environment with RoboCasa/robosuite, SAM2.1, CLIP, and the task assets available.
-
-Real final graph path:
-
-```bash
-MUJOCO_GL=egl PYOPENGL_PLATFORM=egl \
-python examples/realtime_env_graph_eval_loop.py \
-  --task OpenDrawer \
-  --robots PandaOmron \
-  --steps 2 \
-  --visual-backend sam2 \
-  --device cuda \
-  --json-output /tmp/gwam_realtime_final_graph.json
-```
-
-The script calls:
-
-```python
-snapshot = extractor.extract_final_graph(visual_backend=visual_backend)
-graph = snapshot["gnn_graph"]
-```
-
-Expected final-graph invariant:
-
-```text
-D_node = 342
-```
-
-Optional debug snapshot output:
-
-```bash
-python examples/realtime_env_graph_eval_loop.py \
-  --task OpenDrawer \
-  --steps 1 \
-  --visual-backend sam2 \
-  --save-dir /tmp/gwam_realtime_debug \
-  --json-output /tmp/gwam_realtime_debug/summary.json
-```
-
-For CI/docs smoke only, `--visual-backend fake` checks wiring and 342-D shape without SAM2/CLIP. Do not report benchmark metrics from the fake backend.
-
-`--phase1-only-debug` returns the lightweight 33-D Phase-1 graph and is not the final model graph.
-
-## Multi-view RGB / graph alignment contract
-
-For joint training, evaluation, and real-time inference, the model-facing online call is:
-
-```python
-snapshot = extractor.extract_final_graph(visual_backend=visual_backend)
-graph = snapshot["gnn_graph"]          # x [N_real,342], edge_attr [E,8]
-rgb_frames = snapshot["rgb_frames"]             # dict: view_id -> uint8[256,256,3]
-rgb_frame_cameras = snapshot["rgb_frame_cameras"]  # ordered camera names for those view ids
-```
-
-All fields in one `snapshot` are produced from the same current simulator state before the policy/env step advances:
-
-```text
-current qpos/qvel/contact state
-+ current segmentation renders from the ordered camera list
-+ current RGB renders from the same ordered camera list
-+ current visible RLE masks indexed by (node slot n, view id v)
-+ SAM2 masked pooling on rgb_frames[v] using that same mask
-= final graph x [N_real,342]
-```
-
-The ordered multi-view camera contract is:
-
-```text
-v=0 robot0_agentview_right
-v=1 robot0_agentview_left
-v=2 robot0_eye_in_hand
-```
-
-Alignment rules for downstream integration:
-
-- `view_visible[n,v]`, `view_centroid[n,v]`, `view_area[n,v]`, `view_bbox[n,v]`, `rle_masks[*]["v"]`, `visual_features_sparse["v"]`, and `rgb_frames[v]` use the same view id.
-- `visual_features_sparse["n"]` uses the original padded node slot id; `gnn_graph["slot_ids"]` maps active GNN rows back to those padded slots.
-- `rgb_frames[v]` is an upright `uint8[256,256,3]` RGB frame rendered with the same vertical flip convention as the segmentation/mask path.
-- `save_realtime_graph_snapshot(snapshot, out_dir)` writes `rgb/view_<v>_<camera>.npy` plus `rgb_manifest.json`, so saved debug/eval snapshots contain graph data first and aligned RGB images alongside it.
-
-Reusable alignment verifier:
-
-```bash
-MUJOCO_GL=egl PYOPENGL_PLATFORM=egl python scripts/verify_rgb_graph_alignment.py
-```
-
-This verifier checks that one `extract_final_graph()` call does not advance simulator state, that re-rendered current RGB and segmentation/masks match the snapshot, that sparse visual features are reproducible from returned `rgb_frames[v]` and masks, and that saved `rgb_manifest.json`/RGB files round-trip.
-